6 Function Spaces

Data in the real world often takes the form \(y=f(x)+\epsilon \), and a central question is to understand the function space in which \(f\) lives. In this section, we examine several perspectives on function spaces. The section is organized as follows:

1.
Rademacher Complexity
2.
Metric Entropy Method
3.
Glivenko-Cantelli Theorem and Donsker Theorem
4.
Information Theory

This section mainly follows STAT210B (UC Berkeley, taught by Song Mei), High-dimensional probability (PKU, taught by Zhihua Zhang), Introduction to Machine Learning (PKU, taught by Lei Wu), STAT300B (Stanford), STAT364 (Yale). I also referred to the book High-Dimensional Statistics: A Non-Asymptotic Viewpoint [6].

Search definitions, theorems, and topics across the notes.